Capabilities of today's 3D FEM slab design programs 1

[KhouryA]

thread588-217723

I came across the thread “Long span two way PT” and noticed some misconceptions about the capabilities of modern 3D FEM software for the modeling and design of concrete floor systems.

The thread only refers to the capabilities of 2D software in the market and wrongly discounts the capabilities of specialized FEM software for slab design.

Unlike what is posted, slab programs like ADAPT Floor Pro allow for the accurate modeling of complete slab systems using a combination of plate elements for slab regions and frame elements for the beams.

Furthermore, programs like ADAPT Floor Pro calculate cracking and cracked deflections using the actual distribution of reinforcement in the slab.

Khoury

 

[rowingengineer]

Khoury,
Lets open that Can of worms.

"slab programs like ADAPT Floor Pro allow for the accurate modeling of complete slab systems using a combination of plate elements for slab regions and frame elements for the beams."
What happens if you have a wide beam like say 1200mm, how does ADAPT handle this situation? Does it allow you to turn off torsion?

How does ADAPT model columns, are they are a single node connection to the plate elements?

How does ADAPT model walls? Pinned/Fixed/spring?


"Furthermore, programs like ADAPT Floor Pro calculate cracking and cracked deflections using the actual distribution of reinforcement in the slab."

how does ADAPT handle long term deflections? Does it allow you to input the steel directions?

How does ADAPT handle shrinkage curvature in deflections calcs?

I have more questions if you have the time.

Arguing with an engineer is like wrestling with a pig in mud. After a while you realize that them like it

 

[asixth]

Plate elements for wide flat beams, beam elements for deep narrow beams. Does anyone have some good reference material on FE modelling of concrete structures?

 

[rowingengineer]

Asixth,
If you model the beam as a slab, does the program allow for reduced torison stiffness? Does the program allow for beam stiffness ie does the program model teh element in correct elevation?

I don't mean to pull apart hte program, but i am very intrested.

The best book i have come accross is Finite element design of concrete structures: practical problems and their ... By Guenter Axel Rombach

take a look at google books has quite a few pages to review. the slabs section is great for FE programs.

http://books.google.com.au/books?id...e&q=finite element modelling concrete&f=false

Arguing with an engineer is like wrestling with a pig in mud. After a while you realize that them like it

 

[asixth]

I have that text in my collection but have never needed to refer to it. Primarily because I don't get the oppurtunity to design concrete to often and I certainly have never designed a two-way spanning structure using an FE program. It is a topic that I am very interested in and would like to do some further research on in the future.

I have modelled a prestressed bridge with cast insitu slab overlay which I used a frame analysis package to model and did discover that you need a good understanding of the theory before design could take place. For example, I modelled the prestressed units assuming Ig for Ixx because the system was uncracked and reduced J by 80% as per the code. Like the transverse in-situ slab was modelled using 50% Ig for Ixx to account for flexural cracking.

 

[rowingengineer]

Asixth,
there is also a publication by the concrete centre
"How to Design Reinforced Concrete Flat Slabs using Finite Element Analysis" talks about modeling but includes some discussion on the interpretation of results.

Arguing with an engineer is like wrestling with a pig in mud. After a while you realize that them like it

 

[rapt]

All good questions RowingEngineer,

I will answer one of your questions for you, long term effects factor = 2 in Adapt Floor! So if your short term deflection is 0, your total deflection is 0! QED every PT design works without any deflection problems. Too bad ACI318, BS8110, EC2, AS3600 do not allow this approach!
RAPT 2D is far more accurate than that!

More questions for KhouryA:
How can you justify using the results of an FEM slab analysis combined with a banded distributed tendon layout (basically a yieled line design solution)? They are not compatable and can lead to significant underdesign.

How can you justify any design done on the basis of averaging stresses over the full width of a panel, especially after you have gone to the trouble of doing a full FEM analysis to determine the stress dstribution?

How can you justify the inclusion of distribution tendons parallel to band beams in the strength of the band beam as some Adapt software does by default?

And lastly Mxy moments? Most US developed FEM based design software ignores them for design! On what basis?

 

[rowingengineer]

Rapt,
Thanks for the answer, I think guessing the deflection would be better than factoring up the short term deflection.

however you bring up a very interesting point. One FE company has been telling people that AS3600 committee had them do some research on the effect of Mxy on slabs, and from this they were able to advise the committee that it wasn't important and shouldn't e included in AS3600. While I think that Mxy should be included, do you have any info what happened here and what analysis they did, and what conclusions were drawn by the commitee?


Arguing with an engineer is like wrestling with a pig in mud. After a while you realize that them like it

 

[rapt]

RowingEngineer,

I have discussed the design logic for FEM software and Mxy with 3 senior members of the AS3600 committee previously. They are probably the 3 most highly regarded acedemic members of the committee. Their blunt answer was that it must be included and that AS3600 says it must be included (all design actions must be included according to the code) and there was no need to modify AS3600 to be specific about Mxy or they would have then to be specific about every possible design action.

I have also discussed this with several FEM software developers in Europe and several very experienced (in design not just academia) Doctors of Engineering and all agree Mxy must be included in design. None of them would believe that anyone who understood FEM analysis and concrete design would think otherwise. The only people who ignore it are a few USA FEM software developers (as it produces cheaper structures (too bad they might be 15 - 20% understrength)) and therefore means their software gives "better" results than those who do it properly.

I have passed your comment on to the main committee and the chairman of the subcommittee in charge of analysis and will let you know the response in due course, but do not think it will have changed over the last 6 months from the above comments.

PS can you change the "them" in your footer to "they"!

 

[rapt]

RE,

The answer I received back was as I suggested above. Text of the reply is repeated in "" below. If they are saying that they have correspondemnce with the AS3600 committee as you have suggested and AS3600 approval, I would suggest that you name them (at least to AS3600 committee, I will pass it on if you want to contact me privarely) so that the AS3600 committee can have words with them about it. As there are only 2 programs that I know of that ignore Mxy moments in design (at least by default) it is not too hard to guess who they might be!
Maybe they meant that they investigated it for the ACI318 committee, not AS3600. After some dealings with them and knowing some of those involved I would believe it. That is precisely why Australia will NEVER adopt ACI318 while people who understand are making the decisions here.

"Dear Gil,
This is rubbish! I can assure you that neither BD2/1 (main AS3600 committee) nor BD2/5 (analysis and strength sub-committee) have ever asked a software company for such advice, and I am involved in both. Further, it is dangerously incorrect ... at no time are you allowed to violate equilibrium! It is not a committee task or decision to waive a basic principle such as equilibrium ... it is fundamental mechanics.

When I talk to engineers about software packages, I warn them against exactly this case. My advice is, if the software ignores torsions, don’t touch it as the software writers clearly do not understand what it is that they are doing (if they can not get equilibrium correct, what other errors are buried in the software). In the end, it is the engineers responsibility to assess the adequacy of software for the problem at hand and they can not give up this obligation to software writers. It is also certainly not the responsibility of the code committee but the code can (and does) give guiding principles.

All the best
Steve

PS. Torsions are so simply accounted for, I never understood (or understand) why a software package writer would ignore it. Perhaps it says much about some software writers!"

 

[KhouryA]

Contrary to your perception of ADAPT-Floor Pro, this is howit handles the beams (i) if a beam is narrow, user has the option to use a stick element that the program automatically models at the correct eccentricity with respect to the slab, and considers the proper width of the slab at the correct elevation to act with it compositely to resist the applied loads; (ii) user has the option to either fully ignore the torsional rigidity of a beam, or request a user defined fraction of it to be considered in the analysis; (iii) when a slab band is used (you refer to it as a beam with 1200 mm with), user has the option to model it as multiple shell elements to capture the variation of force across the width of the beam – again the slab band will be automatically positions correctly with respect to the flange and interact with it compositely.

Columns can be modeled either as a stick element with a single node connection with the slab or as shell elements extending over the entire width of a column. In either case, the program automatically detects the physical dimension of the columns/slab connection and reports the actions at the faces of the column – a feature not common in other commercially available programs.

Walls are modeled exactly as they appear in their physical condition. Shell elements are used with full capability of defining connection features (fixed, free, etc) to the slab and also at the far ends. The program automatically recognizes the physical condition of a wall/slab connection and attaches a wall either to the soffit or upper surface of a slab. The analytical model created in Floor Pro for a wall is identical to the physical model observed by a viewer. For those who are used to older programs, Floor Pro has provided an option to model supports as springs. But, this is rarely practiced in modern computations.

When calculating cracked deflection, be it immediate or long-term, in addition to program calculated reinforcement, users can define individual bars or reinforcing mesh in length, location and orientation of their choice. Or edit program calculated reinforcement. Floor Pro recognizes the position and orientation of each bar and accounts for its contribution in stiffness of slab/beam at the location of the bar.

In a version of Floor Pro being prepared for release – see their website [

http://www.adaptsoft.com/releases.php

]- the shrinkage curvature at each point is calculated based on (i) user defined exposure of the top and bottom surface, and (ii) the amount, position and orientation of each bar (including prestressing with recognition of bonded and unbonded options). The calculated curvature of each point for user defined time lapse is fed to a full scale 3D finite element model of the floor to determine the resulting three dimensional deflection of the floor system.

 

[rowingengineer]

KhouryA,
Thank you for the reply, appreciated.

Can I ask you to expand on a few area's please.

1.beams, with regards to the line beam, say 600 thick beam, where are the moments for the slab reo design taken from, the edge of the beam or the center of the beam?
With the slab modeled beam can you remove torsion?
2.Columns
how does it handle vertical creep and shrinkage?
Column stiffness how does it computer this, does it take into account the reinforcement in the column and use cracked properties?
3.With regards to the walls fixed vs pinned, can it handle semi-moment transfer? You say u were able to model springs, is this being removed from the software? Does the program do wall stiffness calculations?
4.In the long term defection what method does it us to calculate the long term deflection to take into account creep? Does it take into account time at first loading? Maybe the age adjusted effective modlulus by Ian G?

Again thank you for the reply and time can I now ask a few more questions of you?

a. does the program handle construction sequence load take down?
b. does the program have a way to automatically handle pattern loadings?
c. does the program take into account Mxy when calculating the steel?
d. I noticed that you said the program si using shell elements does is take into account the restraint of walls/column for working out the the tensile forces generated int eh slab by shrinkage and creep?


Arguing with an engineer is like wrestling with a pig in mud. After a while you realize that they like it

 

[rapt]

RE

I recently asked the AS3600 committee for clarification on the design of concrete floors based on FEM analysis and specifically how the Mxy moments should be handled. The following is a copy of their response.As in regard to your technical query we offer the following.

Under no circumstances can torsion be "ignored", it is fundamental to equilibrium!The generally accepted method of dealing with the twisting moments, Mxy, for the design moments is to use the Wood-Armer equations.

In essence, the method involves adding the absolute value of Mxy to the moments Mx and My, using the correct signs of each to give design moments in each direction:
Mx* = Mx + IMxyIM*y = My + IMxyI <?xml:namespace prefix = o ns = "urn:schemas-microsoft-comfficeffice" />

I assume that the context of your question is for the case of slabs. In this case the basic of equation of equilibrium is:

"missing graphic, showing a basic equilibrium equation d2mx/dx2 + d2my/dy2 - 2 * d2mxy/dxdy + q = 0;emphasing that equalibrium must be satisfied and members must be ductile"

Provided this equation is satisfied, equilibrium is satisfied. How we treat this will depend on the method of analysis. For example, in most lower bound methods the torsional component of the stress resultants are taken as zero and the load carried, in full, as moments in the x and y directions (or in the case of a one way slab, in one of the x or y directions). Provided that the stress resultants sum to the total load on the slab, equilibrium is satisfied. Note here that torsion has not been "ignored", rather a concisions decision has been taken by the designer to take mxy as zero and, thus, increase one or both the other components. Further, to ensure that the system has sufficient ductility additional compatibility reinforcement (as per AS3600-2009 clause 9.1.3.3(e)) must be placed in the high torsion regions to alleviate any adverse torsional effects and see that the loads can be redistributed to the designers selected load path.

If we were to use a linear-elastic FE package to obtain the design moments, three bending components (mx, my and mxy) and two shear components (vx and vy) will be output. In this case, the torsional component is likely not zero and most certainly cannot be ignored. This is fundamental mechanics. In this case, the usual method of analysis is to determine the yield condition using the "Wood-Armer" equation:

mux = mx + k|mxy|
muy = my + k-1|mxy|

where k is typically taken as k = 1. Here mux and muy include both the normal and torsional components of the moments.

To "ignore" the torsional moments in such circumstances violates equilibrium and is dangerous!

On the second question, I think that the code is reasonably clear on the torsional stiffness to be taken:

- For equilibrium torsion, the torsional stiffness of the uncracked section should be used.

- For compatibility torsion, the torsional stiffness may be taken as zero provided that "the torsion reinforcement requirements of Clause 8.3.7 and the detailing requirements of Clause 8.3.8 are satisfied." (Refer to Clause 8.3.2).

I trust that this answers your questions; am happy to provided further detail if needed.